matlab - Remove paper texture pattern from a photograph -


I have scanned an old photo with a paper texture pattern and I would like to remove the possible texture as possible. Is there a way to use the image processing toolbox in MATLAB?

I tried to use FFT, but I could not find any clear white spots for painting. Perhaps the pattern is not so regular for this method?

You can see the sample below. If you need the full image, then I can upload it somewhere else. photo sample

Unfortunately, you are stuck in very spatial domains, because the pattern is not really enough to use for foreior analysis.

As told by @Janas and @michid, filtering will help you with this kind of problem. With filtering, you have to face trade-off between the amount of expansion that you want to keep and the amount of noise (or unwanted image components) that you want to remove, for example, to be used by @Jones The average filter paper completely removes the texture (even the round scratch near the bottom edge of the image) but it removes all the textures within the eyes, hair, face and background. Received (although we really care about do not have the background, the foreground is what is important). You will also see a slight decrease in contrast to the image, which is usually undesirable.

  • Find the paper texture pattern:
    • To apply a Gaussian stigma
    • Calculate the image difference between the blurred and the original images
    • Edit Apply the Gaussian blur to the difference image (use a small 3x3 kernel)
    • Threshold using empirically-determined threshold. It generates a binary image that can be used
    • Only

      Paper Texture Pattern (before thresholding):

       Enter image details here

      You want the actual image information present in the image above to be present. You will see that you V Ery brightly slip the edge of the face (this is not good, but this is the best time for me). You also want this paper to be as tangible as possible (so that the thresholding gives the same result across the image). Again, the image above is slightly darker on the right hand, which means that it would be difficult to threshold well.

      Last image:

      Enter image details here

      The result is not correct, but this highly-visible paper texture pattern has been completely removed while preserving more high frequency content than simple filtering approaches.

      Edit

      Filled areas are usually given a plain color and in this way a little bit stands if you look at the image very closely Let's try to add some low-power zero-mean Gaussian noise in the areas of trust so that they can look realistic. You must choose noise variance to match the background. It can be very good to determine it very well.

      There is a process processed with noise:

      Enter image details here

      Note that the parts in which the paper was removed were more difficult to see because the added gaussian noise is masking them. I used the same Gaussian distribution for the whole image but if you want to be more sophisticated then you can use different distribution for face, background, etc.

Comments

Popular posts from this blog

qt - switch/case statement in C++ with a QString type -

python - sqlite3.OperationalError: near "REFERENCES": syntax error - foreign key creating -

Python's equivalent for Ruby's define_method? -